February 4, 2026
According to the consensus of presenters, and most of the people I spoke with, NRF 2026 marked the beginning of a genuinely disruptive phase in retail. Not another cycle of edge-tweaking or rebranding of familiar ideas, but something that felt structurally different.
The NRF (National Retail Federation) conference, known as “Retail’s Big Show”, held in New York each January and attended by around 40,000 people, had an unusual atmosphere this year: a blend of terror and excitement for the future.
With similar views echoed by retailers on the Retail Innovation Council and Retail Tech Show Advisory Board, there is clear optimism about the future. However, hesitancy and uncertainty over where investment should be focused risk slowing progress, while pre-emptive FOMO may also drive rushed decisions made without sufficient strategic intent.
I’ve listed below my retail predictions for 2026, heavily caveated so I can’t possibly be wrong. I also feel the need to stress these are my predictions – not those of a chatbot.
NRF 2026 didn’t reveal a new channel, platform, or vendor category: it revealed a new operating model for commerce.
In the evolving world of interoperable commerce, the buy button is no longer anchored to a channel but embedded directly into conversations, agents, and tools.
For many years, retailers have used phrases such as “Omnichannel” and “Unified Commerce” to describe how customer journeys should be aligned across channels (e.g., web, app, social, stores, etc). At NRF, the phrase “Interoperable Commerce” was used to describe a shift in ownership of those journeys: from channels to surfaces.
If traditional channels, as described above, are designed by retailers to bring customers to them (and hence, maintain stronger control of the customer relationships), then surfaces are locations often outside a brand’s direct control or influence.
Organisational and operating models will need to adapt accordingly; traditional product siloes may limit businesses’ ability to interact with customers away from brand-controlled channels.
Consider changes to organisational design to ensure this rapidly evolving retail model can be accommodated. And from a marketing perspective, consider the differences between SEO, GEO and ACP (SEO finds you, GEO explains you, ACP acts for you).
If retailers can no longer expect to be the customer’s final destination in a transaction (which may now end, for example, in ChatGPT), they need to broadcast their capabilities: what they can do, and what they sell, in machine-readable, API-first format.
Destinations say “come here”. Capabilities say “here’s what I can do for you right now.” For example, product details, pricing, availability, bookings, delivery options, loyalty status, warranties, and more.
Without this, retailers may well exist but agents won’t be able to find them.
Expose capabilities, don’t just build destinations: Work to ensure core services are machine-readable, API-first, and agent-ready.
Google announced the release of the Universal Commerce Protocol (UCP) at NRF. If HTTP made documents interoperable back in 1991, UCP is concerned with the equivalent for transactions. The words “Agent” or “Agentic” are noticeably absent from the protocol’s title, suggesting support for transactions being made across agents, tools and systems.
At the same event, other vendors stressed the importance of agentic technologies, such as commercetools’ “Agentic Commerce Suite”, positioning themselves as the ideal backend to feed a UCP-powered frontend. They highlighted a partnership with Stripe to ensure that when a Google AI agent triggers a UCP checkout, the payment is securely handled by the merchant’s existing commercetools and Stripe setup.
Salesforce further demonstrated its capabilities of their Agentic Commerce Protocol (ACP), which already allows ChatGPT and other OpenAI-based agents to trigger checkouts inside Salesforce Commerce Cloud.
At this point, these technologies don’t feel mutually exclusive, but compatibility matters. Otherwise we risk recreating the extension-heavy ecosystems from which we were supposed to be escaping. Technology departments have challenging decisions to make.
There’s a clear bias toward interoperability: Google CEO Sundar Pichai announced on stage at NRF that Apple will support UCP, following on from Apple’s earlier announcement that Siri will become Gemini-powered.
The emerging battle isn’t between competing protocols, but between genuine interoperability and the gravitational pull of dominant ecosystems. Retailers won’t be locked in by closed standards, but by where the best tooling, data, and economics converge.
Treat protocols as strategic, not technical: Track adoption; push vendors on interoperability; avoid being tempted into single-agent ecosystems. Recognise that UCP or ACP won’t fix legacy chaos.
Merchant of Record was arguably the phrase of the NRF show. In an agentic commerce world, questions remain about impacts on margin, refunds, customer support and tax; and at a macro level, control, liability and trust.
The positive steps of protocols such as Google’s UCP mean, for example, retailers can ensure their name, not Google’s or OpenAI’s, appears on customer bank statements. Whilst this aim is understandably reassuring to retailers, just how much ownership of direct customer relationships needs to be relinquished is not yet fully clear.
Ensure the pull to embrace agentic commerce does not detract from brand values, brand trust and ownership of that customer relationships.
Agentic architectures extend MACH from being predominantly a “sell-side” enabler into the planning, buying and movement layers of retail operations. Historically, the composable principles around MACH – Microservices, API-First, Cloud-Native and Headless have clustered around “sell” since APIs were prevalent and easiest to standardise (e.g.,catalogue, price, basket, checkout, etc). Agentic architecture takes these principles into plan, buy and move.
Although agentic architecture doesn’t magically replace ERPs, Financial Controls, Regulatory Systems and Physical Execution Systems (Warehouses, etc), it repositions and orchestrates them. In this model:
Get systems MACH and Agentic ready: Focus on tech stack modernisation to ensure AI and agentic ambitions aren’t thwarted by legacy systems and processes (such as overnight batch jobs in a real-time world).
Data hygiene is now the minimum cost of entry. If product, inventory, and operational data are locked in legacy systems or inconsistently structured, AI systems simply won’t know a retailer is open for business.
Several NRF startups (e.g. Vody) exist largely to make retail data AI-ready. Strong PIM systems are becoming foundational rather than “nice to have”.
Get data AI-ready: Focus on data housekeeping; clean, standardise, and extract product, inventory, and operational data
Without wishing to repeat a LinkedIn aphorism, AI isn’t replacing humans, but humans powered by AI are coming for those who aren’t. The CEO of Walmart spoke at NRF about how AI is bringing the right information to their people, improving interactions with customers.
AI is increasingly empowering both store colleagues and head-office teams. Vendors such as Enactor demonstrated how business-facing AI prompts can define entire checkout flows, including local tax and fiscal rules, without heavy IT involvement.
Retail has long relied on highly technical business users maintaining critical spreadsheets. AI, combined with modern API-based architectures, now allows those same people to build real systems, not just workarounds, without always “bothering IT”.
Adopt humans + AI as the operating model: Empower colleagues with AI and focus on reducing IT bottlenecks. Historical cultural norms and/or organisational structures may need to be assessed.
Despite the hype, most “agents” on show at NRF demonstrated very low agenticity. We’re still a long way from confidently issuing instructions like:
Agents can execute transactions. Agentic systems decide which transaction should happen, when, and why. Most NRF demos were firmly in the former category.
Handle expectation management: The risk is not over-automation, but over-promising autonomy.
Retail Media continues to grow, with customer context and behaviour acting as strong indicators of intent.
Retailers can increasingly monetise the journey itself, not just the product. Brands will pay to reach high-intent shoppers across websites, apps, social, and physical stores.
Zebra demonstrated in-store handheld devices using proximity and real-time customer data to surface product information, recommendations, and alternatives, described as “hyper-personalised nudges”. Whether customers will embrace messages like “Pepsi is suddenly half-price for you” when scanning a Coke remains to be seen.
Bookings (makeovers, fittings, consultations) are also becoming a powerful signal of intent, linking online and in-store journeys and enabling more targeted retail media.
Monetise the journey as well as the product: Treat intent signals (search, bookings, proximity) as strategic assets, but consider regulatory and trust implications.
The democratisation being enabled by protocols such as UCP means boutique retailers now get to play in the big leagues.
To some extent, UCP solves the “𝑁 x 𝑁 integration problem”. Without such a protocol, 𝑁 retailers would need to build 𝑁 integrations for 𝑁 different AI solutions. UCP reduces this to a 1 x 𝑁 relationship.
In a protocol-driven world, obscurity is no longer a protective moat – it’s a vulnerability.
The most dramatic take I heard at NRF is that we’re witnessing the end of the web; or at least, the end of the web as we know it.
Browser interaction has barely changed since Mosaic in 1993. What has changed is where discovery happens. Nearly 17% of Gen-Z information-seeking now happens in chatbots like Gemini or ChatGPT, compared with ~5% for users over 65. TikTok and Instagram continue to siphon discovery traffic away from Google.
Two post-NRF announcements were sobering:
Plus ça change, as the French may say.
This may be less the “end of the web” and more the end of the web traffic era.
Be realistic about agentic commerce claims. Treat near-term agents as execution-focused, not autonomous, but accept this is likely to be the norm for some time.
In ‘The Fifth Element’, Bruce Willis orders Chinese food and the restaurant literally flies to him, serves the meal, and moves on. It’s a handy metaphor for where retail is heading.
Customers don’t always want to “go shopping”. Often, they want outcomes. And increasingly, those outcomes are being delivered in places retailers don’t own: chat interfaces, agents, feeds, assistants, and tools, exposed via protocols rather than (for example) web pages.
TikTok and Instagram already proved that transactions don’t need websites. NRF 2026 showed what happens next: a convergence toward fewer interaction models, fewer destinations, and more capability-driven commerce.
This shift doesn’t depend on new human behaviour. Intent still predicts action. What’s changed is that retailers can no longer wait for customers to arrive. Commerce now needs to turn up, fully formed, wherever intent appears.
The future won’t belong to the brands with the best websites, but to those whose capabilities are easiest to discover, trust, and execute – by humans and by machines alike.
Within the Retail Technology Service at Equal Experts, we’re working closely with retailers on the points above. Over the next couple of months, we’ll share additional blogs to delve a little deeper into several of the most strategically important topics, particularly agentic commerce. We’ll examine how customer missions (how and why people shop, and how this is rarely consistent) affect your interactions with them, how your technology (stack and department) may need to adjust to accommodate agentic architecture, and how a strategic approach to interoperable commerce can resolve challenges of brand management and customer relationship ownership in an agentic world.
Paul is a Retail Technology Strategy Consultant, who recently joined Equal Experts after spending the last 20+ years working for retailers across the globe, including positions as CTO at Halfords and Chief Architect at New Look, Primark, Marks & Spencer, and Argos in Shanghai.
All business models have their pros and cons. But, when you consider the type of problems we help our clients to solve at Equal Experts, it’s worth thinking about the level of experience and the best consultancy approach to solve them.
If you’d like to find out more about working with us – get in touch. We’d love to hear from you.
